#SD3.5
Explore tagged Tumblr posts
Visit Tumblr Blog
Explore Tumblr blogs with no restrictions, modern design and the best experience.
Last Seen Tumblr Blogs
Fun Fact
In Q3 of 2020, 31% of US users access the Tumblr app daily.
Text
16 notes
路
View notes
Text
馃毃 Big news in AI training stability! Massive strides across multiple architectures, including Flux Dev & Stable Diffusion 3.5. Before/After properly aligning the text encoders. Groundbreaking improvements! More details soon. 馃敟 #AI #DeepLearning #StableDiffusion #FluxDev #AIRevolution #opensource
#stable diffusion#aiartcommunity#flux.1#open source#machine learning#sd3.5#fluxai#artificial intelligence#stability ai#ai training
0 notes
Video
youtube
ComfyUI Face Cropping Workflow for SDXL 3.5M Portraits! 馃帹 Get the workflow here
1 note
路
View note
Text
0 notes
Text
#AWS#Amazon Bedrock#AI#Generative AI#Anthropic Claude 3.7 Sonnet#Anthropic Claude 3.7#Anthropic#Claude 3.7 Sonnet#Claude 3.7#Claude#Stability AI Stable Diffusion 3.5 Large#Stability AI#Stable Diffusion 3.5 Large#Stable Diffusion#SD3.5 Large
0 notes
Text
Fridays AI Art
Art for a new magic custom set. It's in very early stages of design
A few samples from last year. This time last year the new influx of models was starting. SD3, SD3.5 and ultimately Flux. SDXL was still king, but the improvement in prompt adherence was noticeable.
0 notes
Text
I enjoy using Canny with my own drawings because it preserves the line work, which is mine, but fills in the details better than I could; this inspires me to redraw the original using the AI-"enhanced" version as a reference. In this way I can ameliorate the creative atrophy that my current AI obsession represents.
#mine#pencil drawing#sketch#ai generated#stable diffusion#sd3.5#controlnet#canny#line drawing#aunt selma#comfyui
3 notes
路
View notes
Text
This is a very literal interpretation
1 note
路
View note
Text
(Early) Fridays AI ART.
SD3.5 has better stylization than Flux, but worse anatomy.
Still without artist name recognition, it's much harder to get appealing results to my taste.
0 notes
Text
Fridays AI Art
Discussing recent Local AI image generation developments //
Tras el fiasco de SD3M, Stability AI (SAI) ha publicado un nuevo modelo: SD3.5 que puede ser una alternativa interesante a Flux.
Ambos modelos comparten el desconocimiento de nombres de artistas, pero SD3.5, por las pruebas que he hecho, logra un mejor resultado en estilo pict贸rico que Flux. Donde el nuevo modelo no supera a Flux es en anatom铆a y comprensi贸n, no est谩 lejos, pero Flux es superior.
Es curioso tambi茅n, que ahora que SAI no es el 煤nico gallo del gallinero, haya abandonado parte de su mojigater铆a a la hora de entrenar su nuevo modelo.
Si SD3.5 resulta ser m谩s f谩cil de entrenar o si tan solo se comporta mejor cuando se usan Loras, puede ser muy interesante.
Apl铆quese aqu铆 el dicho de nunca es tarde si la dicha es buena, pero es una l谩stima que SAI desperdiciara tanto tiempo trabajando en la seguridad de su modelo, cuando podr铆a haber trabajado en mejorarlo.
0 notes